def test_generate_sub_file(tmpdir):
preds = np.arange(10)
id_sub = np.arange(10)
sub_file_path = tmpdir + 'file.csv'
ms.generate_sub_file(preds, id_sub, sub_file_path)
result_df = DDF.from_csv(sub_file_path)
assert result_df.shape[1] == 2
assert result_df.shape[0] == 10
def test_make_submission_file(tmpdir):
sub_path = str(tmpdir) + 'sub.csv'
model_params = {'min_data': 1, 'min_data_in_bin': 1}
ms.make_submission_file(train_df, test_df, sub_path, extra_model_params=model_params)
result_df = DDF.from_csv(sub_path)
assert 'SalePrice' in result_df.columns
assert 'Id' in result_df.columns
assert len(result_df.columns) == 2
def test_eval_model_on_cv(tmpdir):
df = load_df(TRAIN_PATH, nrows=10)
model_params = {'min_data': 1, 'min_data_in_bin': 1}
log_path = str(tmpdir) + 'file.csv'
ev.eval_model_on_cv(df=df,
log_file=log_path,
extra_model_params=model_params)
results_df = DDF.from_csv(log_path)
assert len(results_df) == 1
assert results_df['rmse'][0] > 0
def test_get_lb_ixs():
df = DDF({
'col1': np.arange(5),
'SalePrice': np.array([100.]*3 + [np.nan]*2)
})
ixs = ms.get_lb_ixs(df)
expected = {
0: {'train': np.array([True]*3 + [False]*2),
'val': np.array([False]*3 + [True]*2)}
}
assert np.all(expected[0]['train'] == ixs[0]['train'])
assert np.all(expected[0]['val'] == ixs[0]['val'])
import numpy as np
from wutils.ddf import DDF
from housePriceDone import make_submission as ms
df = DDF({
'Id': np.arange(3),
'MiscFeature': [np.nan, 'Gar2', 'Othr'],
'MoSold': [2, 5, 9],
'SalePrice': [1000, 20000, 400000]
})
train_df = df.colslice(df.columns)
test_df = df.drop_columns(['SalePrice'])
def test_make_submission_file(tmpdir):
sub_path = str(tmpdir) + 'sub.csv'
model_params = {'min_data': 1, 'min_data_in_bin': 1}
ms.make_submission_file(train_df, test_df, sub_path, extra_model_params=model_params)
result_df = DDF.from_csv(sub_path)
assert 'SalePrice' in result_df.columns
assert 'Id' in result_df.columns
assert len(result_df.columns) == 2
def test_append_dfs():
result_df = ms.append_dfs(train_df, test_df)
assert len(result_df) == len(train_df) + len(test_df)
import numpy as np
from wutils.ddf import DDF
from housePriceDone.data import TRAIN_PATH, load_df
from housePriceDone import eval_model as ev
df = DDF({
'MSZoning': ['FV', 'RH', 'RM'],
'MiscFeature': [np.nan, 'Gar2', 'Othr'],
'MoSold': [2, 5, 9],
'SalePrice': [1000, 20000, 400000]
})
def test_eval_model_on_cv(tmpdir):
df = load_df(TRAIN_PATH, nrows=10)
model_params = {'min_data': 1, 'min_data_in_bin': 1}
log_path = str(tmpdir) + 'file.csv'
ev.eval_model_on_cv(df=df,
log_file=log_path,
extra_model_params=model_params)
results_df = DDF.from_csv(log_path)
assert len(results_df) == 1
assert results_df['rmse'][0] > 0
def test_clean_data():
result_df = ev.clean_data(df)
assert result_df['MSZoning'].dtype == float
assert np.all(df['MoSold'] == result_df['MoSold'])
def load_df(data_path, nrows=None):
return DDF.from_csv(data_path, nrows=nrows)
def generate_sub_file(preds, sub_id, sub_file_path):
df = DDF({
'Id': sub_id,
ev.target_name: np.exp(preds)
})
df.to_csv(sub_file_path)